Fix a minor curiousity with how epochs were counted in the log/progress bar output when training

In the code, as with most Python counters, the epoch (and batch numbers) start from 0. But in the log output this would be somewhat confusing. In particular on the progress bar it would write something like

Epoch 0/5

when running the first epoch. And after running the last epoch would stop on

Epoch 4/5

For the log messages at least, we start the epoch count from 1.